Tracking Welfare Reform: Designing Followup Studies of Recipients Who Leave Welfare. Survey Development


As with all research methods, investigators must decide which questions they are trying to answer. The purpose of the survey will have an impact on its design. Researchers must decide whether they will study applicants, recipients, people who have left welfare, sanctioned individuals, or other groups. They must also decide whether they will use a comparison group. (For example, if a study focuses on people who have left welfare, it might also include a similar group who remained on welfare to identify significant differences between these populations.)

The unit of study, such as individuals, families, households, or children, should be clearly identified. Another consideration is whether to use an independent sample or a panel. With an independent sample, participants are randomly selected for study at each survey point. With a panel, the same individuals are surveyed at each point.

Timing is also crucial. Determining when to ask the survey questions depends on the purpose of the study. Some studies include anyone who received a welfare payment during a particular period; others focus on long-term recipients or former recipients. The timing of the study must also take into account any changes made to the program that might affect outcomes.

The timing of the followup also depends on how long it takes for events to happen; for example, the number of people finding or retaining jobs within one month after leaving welfare is likely to be different from the number employed after six months or one year. The timing also depends on the study design; if the survey will be conducted at one point in time, investigators should allow enough time to pass to capture outcomes. If the survey will be done in stages, it might begin closer to the point that clients leave welfare. Another consideration is that recall deteriorates over time, and experiences, particularly short-term employment, may be inadvertently omitted by respondents.